[SERVER-33570] The logical session collection sync should be majority write Created: 01/Mar/18  Updated: 29/Oct/23  Resolved: 10/Apr/18

Status: Closed
Project: Core Server
Component/s: Internal Code
Affects Version/s: None
Fix Version/s: 3.7.4

Type: Bug Priority: Major - P3
Reporter: Mira Carey Assignee: Matthew Saltz (Inactive)
Resolution: Fixed Votes: 0
Labels: None
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Backwards Compatibility: Fully Compatible
Operating System: ALL
Sprint: Sharding 2018-04-09, Sharding 2018-04-23
Participants:

 Description   

The writes to the logical session collection should be majority writes, so that updates can survive rollbacks.

Currently the reads for the existence of those records are primary required and local. That may be fine, or we may want to do causally consistent reads or majority reads.

We need to ensure that updates stick, but should be fine reading any data after the write we just did. After the initial document insert, any amount of staleness we could see involves rolled back endSessions calls, rolled back deletions from the ttl index or not yet majority written updates to the time stamp. Any of that staleness can only cause you to delay garbage collection, which is acceptable.

We probably also want to set a maxTimeMillis at that point (to avoid perpetual hangs).



 Comments   
Comment by Githook User [ 10/Apr/18 ]

Author:

{'email': 'matthew.saltz@mongodb.com', 'name': 'Matthew Saltz'}

Message: SERVER-33570 Add write concern majority to logical session cache refresh request
Branch: master
https://github.com/mongodb/mongo/commit/3aa87b3580528a673e1338746b6f857a51f47b94

Generated at Thu Feb 08 04:33:51 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.